Expert system for optimization of model reduction techniques

An expert system used for model reduction is described. The system is a useful tool in the heuristic optimization of approximate model design. A set of rules for selection of the appropriate approximation algorithms has been derived from experience. The approach represents a novel way of formalizing rules in order to obtain optimal reduced-order models. The expert system (EXPRED) has been implemented using the M.1 shell of FRAMENTEC, which is one of the most powerful knowledge-based software tools for a personal computer.<
>